def train(): Dtr, Val, Dte = load_data() print('train...') epoch_num = 30 best_model = None min_epochs = 5 min_val_loss = 5 model = cnn().to(device) optimizer = optim.Adam(model.parameters(), lr=0.0008) criterion = nn.CrossEntropyLoss().to(device) # criterion = nn.BCELoss().to(device) for epoch in tqdm(range(epoch_num), ascii=True): train_loss = [] for batch_idx, (data, target) in enumerate(Dtr, 0): try: data, target = Variable(data).to(device), Variable(target.long()).to(device) # target = target.view(target.shape[0], -1) # print(target) optimizer.zero_grad() output = model(data) # print(output) loss = criterion(output, target) loss.backward() optimizer.step() train_loss.append(loss.cpu().item()) except: continue # validation val_loss = get_val_loss(model, Val) model.train() if epoch + 1 > min_epochs and val_loss < min_val_loss: min_val_loss = val_loss best_model = copy.deepcopy(model) torch.save(best_model.state_dict(), r"E:\dataset\Airbnb\training_data\model\cnn.pkl")

UART.zip_reset_uart dtr_uart.h

//DM430-A型开发板串口接收发送程序，使用单片机的串口0，显示操作通过...//注意使用BSL模块作为USB转串口功能的时候，RTS和TCK跳线帽、DTR和RESET跳线帽要断开 //波特率更改请通过config.h文件，直接更改baud参数即可

rs232中断式程序实例.rar_rs232_site:www.pudn.com

2. **串口寄存器**：RS-232通信涉及的主要寄存器包括DTR（Data Terminal Ready）、DSR（Data Set Ready）、RI（Ring Indicator）、DCD（Data Carrier Detect）、RXD（Receive Data）、TXD（Transmit Data）、以及...

TComPort8.zip_VCL Delphi_https://8vcl:com_tcompo_tcomport_vcl

6. **流控制**：支持 RTS/CTS 和 DTR/DSR 等硬件流控制，确保数据传输的正确性和完整性。 7. **线程安全**：在多线程环境中，确保对串口的操作是线程安全的，避免数据冲突。使用 TComPort8 VCL 组件，开发者可以...

这段代码是什么意思from sklearn.neighbors import KNeighborsClassifier from sklearn.datasets import load_digits from sklearn.model_selection import train_test_split x=wine_data.iloc[:,1:].values y=wine_data.iloc[:,0].values x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,random_state=125) dtr=KNeighborsClassifier() dtr.fit(x_train,y_train) dtr.score(x_test,y_test)

接着，使用train_test_split函数将数据集分为训练集和测试集，其中test_size=0.3表示测试集占总数据集的30%，random_state=125是为了保证每次运行代码时得到的训练集和测试集是固定的。然后，创建一个...

def load_data(file_name): df = pd.read_csv('data/new_data/' + file_name, encoding='gbk') columns = df.columns df.fillna(df.mean(), inplace=True) return df class MyDataset(Dataset): def init(self, data): self.data = data def getitem(self, item): return self.data[item] def len(self): return len(self.data) def nn_seq_us(B): print('data processing...') dataset = load_data() # split train = dataset[:int(len(dataset) * 0.6)] val = dataset[int(len(dataset) * 0.6):int(len(dataset) * 0.8)] test = dataset[int(len(dataset) * 0.8):len(dataset)] m, n = np.max(train[train.columns[1]]), np.min(train[train.columns[1]]) def process(data, batch_size): load = data[data.columns[1]] load = load.tolist() data = data.values.tolist() load = (load - n) / (m - n) seq = [] for i in range(len(data) - 24): train_seq = [] train_label = [] for j in range(i, i + 24): x = [load[j]] train_seq.append(x) # for c in range(2, 8): # train_seq.append(data[i + 24][c]) train_label.append(load[i + 24]) train_seq = torch.FloatTensor(train_seq) train_label = torch.FloatTensor(train_label).view(-1) seq.append((train_seq, train_label)) # print(seq[-1]) seq = MyDataset(seq) seq = DataLoader(dataset=seq, batch_size=batch_size, shuffle=False, num_workers=0, drop_last=True) return seq Dtr = process(train, B) Val = process(val, B) Dte = process(test, B) return Dtr, Val, Dte, m, n这是关于什么的代码

代码中的 load_data 函数用于加载数据文件，MyDataset 类用于创建数据集对象，nn_seq_us 函数则用于对数据进行预处理并将其转换为 PyTorch 数据加载器对象。最后，该函数返回训练、验证和测试数据集对象，以及数据的...

wine_data=data.iloc[:-5,:] wine_target=data.iloc[-5:,:] from sklearn.tree import DecisionTreeClassifier from sklearn.model_selection import train_test_split from sklearn.linear_model import LogisticRegression from sklearn.model_selection import train_test_split x=wine_data.iloc[:,1:].values y=wine_data.iloc[:,0].values x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,random_state=42) #建立模型 dtc=DecisionTreeClassifier(criterion='entropy')#基于熵评价纯度 dtc.fit(x_train,y_train)#拟合数据 y_pre=dtc.predict(x_test) y_pre dtc.predict(wine_target.iloc[:,1:].values) from sklearn.metrics import mean_squared_error #先获得预测的y值y_pre _pre=dtc.Oredlct(y tact mean_squared_error(y_test,y_pre) print("决策树训练精度：“,dtc.score(x_test,y_test)） print("决策树泛化精度：“,dtc.score(x_train,y_train)) #KNN最近邻分类算法 from sklearn.neighbors import KNeighborsClassifier from sklearn.datasets import load_digits from sklearn.model_selection import train_test_split -wine_data.ilocl:,1:].values y=wine_data.iloc[:,0].values state=125） dtr=KNeighborsClassifier() dtr.fit(x_train,y_train) dtr.score(x_test,y_test) model_knn=KNeighborsClassifier(n_neighbors=5)#看5个近邻的类别确定分类 model knn.fit(x_train,y_train) #预测 model_knn.predict(x_test) dtr.predict(wine_target.iloc[:,1:].values) neighbors = 3 from sklearn.neighbors import KNeighborsClassifier knn = KNeighborsClassifier(neighbors) knn.fit(x_train,y_train) print("KNN 训练精度：“,knn.score(x_test,y_test)) print("KNN泛化精度： knn.score(x_train,y_train))代码解释

接着，使用train_test_split函数将wine_data划分为训练集和测试集，其中测试集占总样本的30%。其次，使用DecisionTreeClassifier类来实现决策树分类算法，其中criterion='entropy'表示使用熵来评价纯度。使用fit...

In [16]: wine_data=data.iloc[:-5,:] wine_target=data.iloc[-5:,:] In [17]: from sklearn.tree import DecisionTreeClassifier from sklearn.model_selection import train_test_split from sklearn.linear_model import LogisticRegression from sklearn.model_selection import train_test_split x=wine_data.iloc[:,1:].values y=wine_data.iloc[:,0].values x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,random_state=42) #建立模型 dtc=DecisionTreeClassifier(criterion='entropy')#基于熵评价纯度 dtc.fit(x_train,y_train)#拟合数据 y_pre=dtc.predict(x_test) y_pre Out[17]: array([3.0, 1.0, 3.0, 2.0, 2.0, 2.0, 2.0, 1.0, 3.0, 2.0, 3.0, 1.0, 2.0, 3.0, 2.0, 1.0, 2.0, 1.0, 3.0, 2.0, 2.0, 2.0, 2.0, 1.0, 3.0, 2.0, 3.0, 1.0, 2.0, 1.0, 2.0, 1.0, 2.0, 2.0, 3.0, 3.0, 2.0, 1.0, 1.0, 1.0, 1.0, 1.0, 3.0, 1.0, 1.0, 1.0, 2.0, 1.0, 2.0, 1.0, 3.0, 3.0]) In [18]: dtc.predict(wine_target.iloc[:,1:].values) Out[18]: array([2.0, 2.0, 2.0, 3.0, 1.0]) In [19]: from sklearn.metrics import mean_squared_error #先获得预测的y值y_pre y_pre=dtc.predict(x_test) mean_squared_error(y_test,y_pre) Out[19]: 0.0 In [20]: print("决策树训练精度：",dtc.score(x_test,y_test)) print("决策树泛化精度：",dtc.score(x_train,y_train)) 决策树训练精度： 1.0 决策树泛化精度： 1.0 In [21]: #KNN最近邻分类算法 from sklearn.neighbors import KNeighborsClassifier from sklearn.datasets import load_digits from sklearn.model_selection import train_test_split x=wine_data.iloc[:,1:].values y=wine_data.iloc[:,0].values x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,random_state=125) dtr=KNeighborsClassifier() dtr.fit(x_train,y_train) dtr.score(x_test,y_test) Out[21]: 0.9807692307692307 In [22]: model_knn=KNeighborsClassifier(n_neighbors=5)#看5个近邻的类别确定分类 model_knn.fit(x_train,y_train) #预测 model_knn.predict(x_test) Out[22]: array([3.0, 3.0, 1.0, 2.0, 1.0, 3.0, 3.0, 1.0, 2.0, 3.0, 2.0, 1.0, 2.0, 1.0, 2.0, 1.0, 2.0, 2.0, 3.0, 1.0, 1.0, 3.0, 1.0, 2.0, 1.0, 2.0, 3.0, 3.0, 2.0, 2.0, 1.0, 1.0, 2.0, 1.0, 1.0, 2.0, 3.0, 1.0, 3.0, 3.0, 2.0, 2.0, 2.0, 2.0, 1.0, 1.0, 1.0, 1.0, 2.0, 3.0, 2.0, 1.0]) In [23]: dtr.predict(wine_target.iloc[:,1:].values) Out[23]: array([2.0, 2.0, 3.0, 3.0, 1.0]) In [24]: neighbors = 3 from sklearn.neighbors import KNeighborsClassifier knn = KNeighborsClassifier(neighbors) knn.fit(x_train,y_train) print("KNN 训练精度：",knn.score(x_test,y_test)) print("KNN 泛化精度：",knn.score(x_train,y_train)) KNN 训练精度： 0.9615384615384616 KNN 泛化精度： 0.9586776859504132代码解释

这段代码主要是进行机器学习的分类任务，使用了决策树和K近邻算法。首先将数据分为训练集和测试集，然后使用DecisionTreeClassifier和KNeighborsClassifier分别建立决策树和K近邻分类模型，并使用训练数据拟合模型。...

import pandas as pd import numpy as np import matplotlib.pyplot as plt import seaborn as sns from sklearn.model_selection import train_test_split, RepeatedKFold, cross_val_score from sklearn.ensemble import GradientBoostingRegressor, RandomForestRegressor from sklearn.svm import SVR from sklearn.tree import DecisionTreeRegressor from sklearn.gaussian_process import GaussianProcessRegressor from sklearn.gaussian_process.kernels import RBF, Matern from sklearn.metrics import mean_absolute_error, r2_score from sklearn.preprocessing import StandardScaler from sklearn.pipeline import make_pipeline from bayes_opt import BayesianOptimization import warnings # 忽略警告 warnings.filterwarnings('ignore') # 设置中文字体支持 plt.rcParams['font.sans-serif'] = ['Arial', 'Arial Unicode MS', 'Microsoft YaHei', 'sans-serif'] plt.rcParams['axes.unicode_minus'] = False sns.set_style("whitegrid") # 加载数据集 df = pd.read_csv('/Jupyter/BCC_HEA/alloy_properties.csv') features = ['mean_r', 'mean_electronegativity', 'mean_G', 'mean_vec', 'mean_delta','mean_delta_G'] targets = ['deta_E_mono','deta_E_di'] # 定义模型名称列表 model_names = ['GBR', 'SVR', 'DTR', 'RFR', 'GPR'] # 为每个目标变量进行操作 for target in targets: print(f"\n\n{'='50}") print(f"开始处理目标变量: {target}") print(f"{'='50}\n") # 数据拆分：训练集+验证集+测试集 X = df[features] y = df[target] # 拆分测试集（30%） X_train_full, X_test, y_train_full, y_test = train_test_split(X, y, test_size=0.3, random_state=42) # 从训练集拆分验证集（20%用于调参） X_train, X_val, y_train, y_val = train_test_split(X_train_full, y_train_full, test_size=0.2, random_state=42) # 存储模型性能 model_performance_before = {} model_performance_after = {} best_params_dict = {} # 1. 初始模型训练和评估（测试集指标为主） print(f"\n{'='30} {target} - 原始模型性能 {'='30}") # 定义初始模型 models = { 'GBR': GradientBoostingRegressor(random_state=42), 'SVR': make_pipeline(StandardScaler(), SVR(kernel='rbf')), 'DTR': DecisionTreeRegressor(random_state=42), 'RFR': RandomForestRegressor(random_state=42, n_jobs=-1), 'GPR': make_pipeline(StandardScaler(), GaussianProcessRegressor(random_state=42)) } # 打印原始模型的默认超参数 print(f'\n\n===== {target} - 原始模型的默认超参数 =====') for model_name, model in models.items(): print(f"\n{model_name} 的默认超参数:") default_params = model.get_params() for param, value in default_params.items(): print(f" {param}: {value}") # 训练和评估初始模型（测试集指标） for model_name in model_names: model = models[model_name] model.fit(X_train_full, y_train_full) # 完整训练集训练 # 预测 y_train_pred = model.predict(X_train_full) # 训练集预测 y_test_pred = model.predict(X_test) # 测试集预测 # 计算指标 train_mae = mean_absolute_error(y_train_full, y_train_pred) train_r2 = r2_score(y_train_full, y_train_pred) test_mae = mean_absolute_error(y_test, y_test_pred) test_r2 = r2_score(y_test, y_test_pred) # 交叉验证（可选，辅助参考） cv = RepeatedKFold(n_splits=5, n_repeats=3, random_state=42) cv_score = cross_val_score(model, X_train_full, y_train_full, cv=cv, scoring='neg_mean_absolute_error').mean() cv_mae = -cv_score # 存储性能 model_performance_before[model_name] = { 'Train MAE': train_mae, 'Train R2': train_r2, 'Test MAE': test_mae, 'Test R2': test_r2, 'CV MAE': cv_mae } # 打印结果 print(f"\n{model_name} 原始模型性能:") print(f" 训练集: MAE = {train_mae:.4f}, R² = {train_r2:.4f}") print(f" 测试集: MAE = {test_mae:.4f}, R² = {test_r2:.4f}") print(f" 交叉验证 MAE = {cv_mae:.4f}") # 绘制初始拟合图（训练+测试集） plt.figure(figsize=(8, 6)) plt.scatter(y_train_full, y_train_pred, alpha=0.7, label='training set', c='blue') plt.scatter(y_test, y_test_pred, alpha=0.7, label='test set', c='red') # 趋势线 all_y = np.concatenate([y_train_full, y_test]) all_pred = np.concatenate([y_train_pred, y_test_pred]) z = np.polyfit(all_y, all_pred, 1) p = np.poly1d(z) plt.plot(all_y, p(all_y), c='black', lw=2, linestyle='--') plt.xlabel('actual', fontsize=12) plt.ylabel('predicted', fontsize=12) plt.title(f'{target} - {model_name} (before optimization)', fontsize=14) # 标注指标 plt.text(0.05, 0.9, f'training MAE = {train_mae:.3f}, $R^2$ = {train_r2:.3f}', transform=plt.gca().transAxes, fontsize=10, bbox=dict(facecolor='white', alpha=0.8)) plt.text(0.05, 0.8, f'test set MAE = {test_mae:.3f}, $R^2$ = {test_r2:.3f}', transform=plt.gca().transAxes, fontsize=10, bbox=dict(facecolor='white', alpha=0.8)) plt.legend(fontsize=10) plt.tight_layout() plt.savefig(f'{target}_{model_name}_before_optimization.png', dpi=300) plt.show() # 2. 贝叶斯优化（验证集MAE为目标） print(f"\n{'='30} {target} - 贝叶斯优化模型超参数 {'='30}") # 超参数空间（保持原范围） pbounds_dict = { 'GBR': { 'n_estimators': (50, 500), 'learning_rate': (0.001, 0.3), 'max_depth': (2, 10), 'min_samples_split': (2, 20), 'min_samples_leaf': (1, 10) }, 'SVR': { 'logC': (0, 3), # C: 1-1000 'logGamma': (-3, 1), # gamma: 0.001-10 'epsilon': (0.01, 0.1) # ε范围 }, 'DTR': { 'max_depth': (2, 15), 'min_samples_split': (2, 20), 'min_samples_leaf': (1, 15), 'max_features': (0.3, 1.0) }, 'RFR': { 'n_estimators': (50, 300), 'max_depth': (3, 10), 'min_samples_split': (3, 8), 'min_samples_leaf': (1, 5), 'max_features': (0.5, 1.0) }, 'GPR': { 'kernel_type': (0, 1), # 0=RBF, 1=Matern 'log_length_scale': (-1, 2), # length_scale: 0.1-100 'log_alpha': (-3.5, -1.5), # alpha: 0.00001-1 'nu': (0.5, 2) # Matern核平滑度 } } # 目标函数：验证集MAE（最小化MAE → 最大化负MAE） def get_objective_function(model_name): def objective_function(params): # 构建模型 if model_name == 'GBR': model = GradientBoostingRegressor( n_estimators=int(params['n_estimators']), learning_rate=params['learning_rate'], max_depth=int(params['max_depth']), min_samples_split=int(params['min_samples_split']), min_samples_leaf=int(params['min_samples_leaf']), random_state=42 ) elif model_name == 'SVR': C_val = 10 params['logC'] gamma_val = 10 params['logGamma'] model = make_pipeline( StandardScaler(), SVR(kernel='rbf', C=C_val, gamma=gamma_val, epsilon=params['epsilon']) ) elif model_name == 'DTR': model = DecisionTreeRegressor( max_depth=int(params['max_depth']), min_samples_split=int(params['min_samples_split']), min_samples_leaf=int(params['min_samples_leaf']), max_features=params['max_features'], random_state=42 ) elif model_name == 'RFR': model = RandomForestRegressor( n_estimators=int(params['n_estimators']), max_depth=int(params['max_depth']), min_samples_split=int(params['min_samples_split']), min_samples_leaf=int(params['min_samples_leaf']), max_features=params['max_features'], random_state=42, n_jobs=-1 ) elif model_name == 'GPR': kernel_type = int(round(params['kernel_type'])) length_scale = 10 params['log_length_scale'] alpha_val = 10 params['log_alpha'] if kernel_type == 0: kernel = RBF(length_scale=length_scale) else: kernel = Matern( length_scale=length_scale, nu=params['nu'] ) model = make_pipeline( StandardScaler(), GaussianProcessRegressor( kernel=kernel, alpha=alpha_val, n_restarts_optimizer=5, random_state=42 ) ) # 训练（用训练集）+ 验证（用验证集） model.fit(X_train, y_train) y_val_pred = model.predict(X_val) val_mae = mean_absolute_error(y_val, y_val_pred) return -val_mae # 最大化负MAE等价于最小化MAE return objective_function # 优化每个模型 for model_name in model_names: print(f"\n>>> 正在优化 {model_name} 模型...") optimizer = BayesianOptimization( f=get_objective_function(model_name), pbounds=pbounds_dict[model_name], random_state=42, allow_duplicate_points=True ) # 执行优化（增加迭代次数提升搜索充分性） optimizer.maximize( init_points=15, # 初始随机点从10→15 n_iter=30 # 迭代次数从20→30 ) # 获取最优参数 best_params = optimizer.max['params'] best_val_mae = -optimizer.max['target'] # 还原为MAE值 best_params_dict[model_name] = best_params print(f"\n{model_name} 最优超参数:") # 特殊参数转换显示 if model_name == 'SVR': print(f" C: {10best_params['logC']:.4f} (logC: {best_params['logC']:.4f})") print(f" gamma: {10best_params['logGamma']:.6f} (logGamma: {best_params['logGamma']:.4f})") print(f" epsilon: {best_params['epsilon']:.4f}") elif model_name == 'GPR': kernel_type = "RBF" if int(round(best_params['kernel_type'])) == 0 else "Matern" print(f" 核类型: {kernel_type}") print(f" length_scale: {10best_params['log_length_scale']:.4f} (log: {best_params['log_length_scale']:.4f})") print(f" alpha: {10**best_params['log_alpha']:.6f} (log: {best_params['log_alpha']:.4f})") if kernel_type == "Matern": print(f" nu: {best_params['nu']:.4f}") else: for param, value in best_params.items(): if 'int' in str(type(value)) or param in ['n_estimators', 'max_depth', 'min_samples_split', 'min_samples_leaf']: print(f" {param}: {int(value)}") else: print(f" {param}: {value:.6f}") print(f"最优验证集 MAE: {best_val_mae:.4f}") # 3. 优化后模型评估（测试集指标为主） print(f"\n{'='30} {target} - 优化后模型性能 {'='30}") for model_name in model_names: best_params = best_params_dict[model_name] # 初始化优化后模型（同原逻辑） if model_name == 'GBR': optimized_model = GradientBoostingRegressor( n_estimators=int(best_params['n_estimators']), learning_rate=best_params['learning_rate'], max_depth=int(best_params['max_depth']), min_samples_split=int(best_params['min_samples_split']), min_samples_leaf=int(best_params['min_samples_leaf']), random_state=42 ) elif model_name == 'SVR': C_val = 10 best_params['logC'] gamma_val = 10 best_params['logGamma'] optimized_model = make_pipeline( StandardScaler(), SVR(kernel='rbf', C=C_val, gamma=gamma_val, epsilon=best_params['epsilon']) ) elif model_name == 'DTR': optimized_model = DecisionTreeRegressor( max_depth=int(best_params['max_depth']), min_samples_split=int(best_params['min_samples_split']), min_samples_leaf=int(best_params['min_samples_leaf']), max_features=best_params['max_features'], random_state=42 ) elif model_name == 'RFR': optimized_model = RandomForestRegressor( n_estimators=int(best_params['n_estimators']), max_depth=int(best_params['max_depth']), min_samples_split=int(best_params['min_samples_split']), min_samples_leaf=int(best_params['min_samples_leaf']), max_features=best_params['max_features'], random_state=42, n_jobs=-1 ) elif model_name == 'GPR': kernel_type = int(round(best_params['kernel_type'])) length_scale = 10 best_params['log_length_scale'] alpha_val = 10 best_params['log_alpha'] if kernel_type == 0: kernel = RBF(length_scale=length_scale) else: kernel = Matern( length_scale=length_scale, nu=best_params['nu'] ) optimized_model = make_pipeline( StandardScaler(), GaussianProcessRegressor( kernel=kernel, alpha=alpha_val, n_restarts_optimizer=5, random_state=42 ) ) # 训练：用完整训练集（训练+验证） optimized_model.fit(X_train_full, y_train_full) # 预测测试集 y_test_pred = optimized_model.predict(X_test) # 预测训练集（可选） y_train_full_pred = optimized_model.predict(X_train_full) # 计算指标 train_mae = mean_absolute_error(y_train_full, y_train_full_pred) train_r2 = r2_score(y_train_full, y_train_full_pred) test_mae = mean_absolute_error(y_test, y_test_pred) test_r2 = r2_score(y_test, y_test_pred) # 交叉验证（可选） cv = RepeatedKFold(n_splits=5, n_repeats=3, random_state=42) cv_score = cross_val_score(optimized_model, X_train_full, y_train_full, cv=cv, scoring='neg_mean_absolute_error').mean() cv_mae = -cv_score # 存储性能 model_performance_after[model_name] = { 'Train MAE': train_mae, 'Train R2': train_r2, 'Test MAE': test_mae, 'Test R2': test_r2, 'CV MAE': cv_mae } # 打印结果 print(f"\n{model_name} 优化后模型性能:") print(f" 训练集: MAE = {train_mae:.4f}, R² = {train_r2:.4f}") print(f" 测试集: MAE = {test_mae:.4f}, R² = {test_r2:.4f}") print(f" 交叉验证 MAE = {cv_mae:.4f}") # 绘制优化后拟合图（训练+测试集） plt.figure(figsize=(8, 6)) plt.scatter(y_train_full, y_train_full_pred, alpha=0.7, label='training set', c='blue') plt.scatter(y_test, y_test_pred, alpha=0.7, label='test set', c='red') # 趋势线 all_y = np.concatenate([y_train_full, y_test]) all_pred = np.concatenate([y_train_full_pred, y_test_pred]) z = np.polyfit(all_y, all_pred, 1) p = np.poly1d(z) plt.plot(all_y, p(all_y), c='black', lw=2, linestyle='--') plt.xlabel('actual', fontsize=12) plt.ylabel('predicted', fontsize=12) plt.title(f'{target} - {model_name} (after optimization)', fontsize=14) # 标注指标 plt.text(0.05, 0.9, f'training MAE = {train_mae:.3f}, $R^2$ = {train_r2:.3f}', transform=plt.gca().transAxes, fontsize=10, bbox=dict(facecolor='white', alpha=0.8)) plt.text(0.05, 0.8, f'test set MAE = {test_mae:.3f}, $R^2$ = {test_r2:.3f}', transform=plt.gca().transAxes, fontsize=10, bbox=dict(facecolor='white', alpha=0.8)) plt.legend(fontsize=10) plt.tight_layout() plt.savefig(f'{target}_{model_name}_after_optimization.png', dpi=300) plt.show() # 4. 性能对比分析（测试集指标为主） print(f"\n{'='30} {target} - 优化前后模型性能对比 {'='30}") # 构建对比数据 comparison_data = [] for model_name in model_names: before = model_performance_before[model_name] after = model_performance_after[model_name] # 计算提升率 mae_test_improve = (before['Test MAE'] - after['Test MAE']) / before['Test MAE'] * 100 if before['Test MAE'] != 0 else 0 r2_test_improve = (after['Test R2'] - before['Test R2']) * 100 # 百分点 comparison_data.append({ '模型': model_name, '优化前测试MAE': before['Test MAE'], '优化后测试MAE': after['Test MAE'], 'MAE提升(%)': mae_test_improve, '优化前测试R²': before['Test R2'], '优化后测试R²': after['Test R2'], 'R²提升(百分点)': r2_test_improve, '优化前交叉验证MAE': before['CV MAE'], '优化后交叉验证MAE': after['CV MAE'] }) comparison_df = pd.DataFrame(comparison_data) print("\n模型性能对比表:") print(comparison_df.round(4)) # 保存对比结果 comparison_df.to_csv(f'model_comparison_{target}.csv', index=False) # 绘制测试集MAE对比 plt.figure(figsize=(12, 8)) width = 0.35 x = np.arange(len(model_names)) plt.bar(x - width/2, comparison_df['优化前测试MAE'], width, label='before optimization', color='skyblue') plt.bar(x + width/2, comparison_df['优化后测试MAE'], width, label='after optimization', color='lightcoral') # 数据标签 for i, (before, after) in enumerate(zip(comparison_df['优化前测试MAE'], comparison_df['优化后测试MAE'])): plt.text(i - width/2, before + 0.005, f'{before:.4f}', ha='center') plt.text(i + width/2, after + 0.005, f'{after:.4f}', ha='center') # 提升百分比 improvement = (before - after) / before * 100 if before != 0 else 0 plt.text(i, max(before, after) + 0.01, f'{improvement:.1f}%', ha='center', fontsize=10, color='green' if improvement > 0 else 'red') plt.xlabel('Model', fontsize=12) plt.ylabel('test set MAE', fontsize=12) plt.title(f'{target} - before/after optimization test set MAE comparison', fontsize=14) plt.xticks(x, model_names) plt.legend() plt.tight_layout() plt.savefig(f'{target}_MAE_comparison.png', dpi=300) plt.show() # 绘制测试集R²对比 plt.figure(figsize=(12, 8)) plt.bar(x - width/2, comparison_df['优化前测试R²'], width, label='before optimization', color='skyblue') plt.bar(x + width/2, comparison_df['优化后测试R²'], width, label='after optimization', color='lightcoral') # 数据标签 for i, (before, after) in enumerate(zip(comparison_df['优化前测试R²'], comparison_df['优化后测试R²'])): plt.text(i - width/2, before + 0.01, f'{before:.4f}', ha='center') plt.text(i + width/2, after + 0.01, f'{after:.4f}', ha='center') # 提升百分点 improvement = (after - before) * 100 plt.text(i, max(before, after) + 0.02, f'+{improvement:.1f}pp' if improvement > 0 else f'{improvement:.1f}pp', ha='center', fontsize=10, color='green' if improvement > 0 else 'red') plt.xlabel('Model', fontsize=12) plt.ylabel('test set R²', fontsize=12) plt.title(f'{target} - before/after optimization test set R² comparison', fontsize=14) plt.xticks(x, model_names) plt.legend() plt.tight_layout() plt.savefig(f'{target}_R2_comparison.png', dpi=300) plt.show() # 交叉验证MAE对比（可选） all_cv_scores = [] for model_name in model_names: before_score = model_performance_before[model_name]['CV MAE'] after_score = model_performance_after[model_name]['CV MAE'] all_cv_scores.append([model_name + '_before', before_score]) all_cv_scores.append([model_name + '_after', after_score]) cv_df = pd.DataFrame(all_cv_scores, columns=['Model', 'CV MAE']) plt.figure(figsize=(12, 6)) sns.boxplot(x='Model', y='CV MAE', data=cv_df) plt.title('Cross-Validation MAE Comparison Before and After Optimization') plt.ylabel('MAE') plt.tight_layout() plt.savefig(f'{target}_cv_comparison.png', dpi=300) plt.show() print("\n所有模型优化完成！")修改上述贝叶斯优化模型前后的结果比较，不适用交叉验证

根据用户要求，我们需要修改代码，... print(f"\n{'='*30} {target} - 原始模型性能 {'='*30}") # ...（模型定义和参数打印保持不变）... # 训练和评估初始模型（移除交叉验证） for model_name in model_names:

dtc.predict(wine_target.iloc[:,1:].values) Out[33]: array([2., 2., 2., 3., 1.]) In [34]: from sklearn.metrics import mean_squared_error #先获得预测的y值y_pre y_pre=dtc.predict(x_test) mean_squared_error(y_test,y_pre) Out[34]: 0.0 In [35]: print("决策树训练精度：",dtc.score(x_test,y_test)) print("决策树泛化精度：",dtc.score(x_train,y_train)) 决策树训练精度： 1.0 决策树泛化精度： 1.0 In [39]: #KNN最近邻分类算法 from sklearn.neighbors import KNeighborsClassifier from sklearn.datasets import load_digits from sklearn.model_selection import train_test_split x=wine_data.iloc[:,1:].values y=wine_data.iloc[:,0].values x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,random_state=125) dtr=KNeighborsClassifier() dtr.fit(x_train,y_train) dtr.score(x_test,y_test) Out[39]: 0.9807692307692307 In [42]: model_knn=KNeighborsClassifier(n_neighbors=5)#看5个近邻的类别确定分类 model_knn.fit(x_train,y_train) #预测 model_knn.predict(x_test) Out[42]: array([3., 3., 1., 2., 1., 3., 3., 1., 2., 3., 2., 1., 2., 1., 2., 1., 2., 2., 3., 1., 1., 3., 1., 2., 1., 2., 3., 3., 2., 2., 1., 1., 2., 1., 1., 2., 3., 1., 3., 3., 2., 2., 2., 2., 1., 1., 1., 1., 2., 3., 2., 1.]) In [43]: dtr.predict(wine_target.iloc[:,1:].values) Out[43]: array([2., 2., 3., 3., 1.]) In [41]: neighbors = 3 from sklearn.neighbors import KNeighborsClassifier knn = KNeighborsClassifier(neighbors) knn.fit(x_train,y_train) print("KNN 训练精度：",knn.score(x_test,y_test)) print("KNN 泛化精度：",knn.score(x_train,y_train)) KNN 训练精度： 0.9615384615384616 KNN 泛化精度： 0.9586776859504132

这段代码主要是使用决策树分类器和K最近邻分类器对葡萄酒数据进行分类，其中，训练集和测试集的划分、分类器的建立和训练、模型评估等步骤都有详细的代码实现。最终，通过输出score值和mean_squared_error值，可以...

相关推荐

UART.zip_reset_uart dtr_uart.h

rs232中断式程序实例.rar_rs232_site:www.pudn.com

TComPort8.zip_VCL Delphi_https://8vcl:com_tcompo_tcomport_vcl

kiengremsy:stm32flash_bootloader_dtr_rts

Com_RTS_DTR.zip_RTS DTR 下载_VC控制_com dtr_dtr串口_串口RTS

UART__RS232_的CTS与RTS.pdf

PCI_PTS_PO__DTRs_v5-1.pdf

WinSoft_ComPort.NET_v3.0_Downloadly.ir.rar

PL2303_Prolific_WINDOWS_DriverInstaller_v1.14.0.rar

191004_fbo_ggs_Linux_x64_shiphome.zip

造纸机变频分布传动与Modbus RTU通讯技术的应用及其实现

langchain4j-neo4j-0.29.1.jar中文文档.zip

大家在看

ansible-role-kubernetes：Ansible角色-Kubernetes

volume-visualization

波特率任意设 串口调试助手

AIPEX练习手册

爬取招行外汇网站数据.pdf

最新推荐

基于C语言的RS232串行接口通信实现_葛磊蛟.pdf

造纸机变频分布传动与Modbus RTU通讯技术的应用及其实现

langchain4j-neo4j-0.29.1.jar中文文档.zip

基于STC89C52单片机的智能衣架电路设计：服装店顾客行为数据分析与传输

Visual C++.NET编程技术实战指南

HarmonyOS内核深度探秘：优化自由行旅游系统的策略

tkinter模块所有控件

局域网五子棋游戏：娱乐与聊天的完美结合

自由行旅游新篇章：HarmonyOS技术融合与系统架构深度解析

足底支撑相到达73%是什么问题

波特率任意设串口调试助手